Deep learning-based organoid quality control method, storage medium

Abstract

The application provides a kind of organoid quality control method based on deep learning, storage medium, method includes: in pre-drug organoid microscopic image and post-drug organoid microscopic image, each organoid is framed out;A model including convolutional neural network and multi-task learning neural network is constructed;According to the pre-drug organoid microscopic image and post-drug organoid microscopic image of each organoid framed out, the convolutional neural network is trained based on anchor strategy, so that the convolutional neural network has the ability to identify each organoid from image;According to the effective feature that has discriminative to organoid quality evaluation, the multi-task learning neural network is trained, so that the multi-task learning neural network has the ability to evaluate organoid development quality;The evaluation model obtained after training the model is obtained.The application can avoid the disadvantage that ATP method is sensitive to determination time, improve the accuracy of organoid quality evaluation.

Description

Technical Field

[0001] This invention relates to the field of cell quality monitoring technology, and in particular to a deep learning-based organoid quality control method and a computer-readable storage medium. Background Technology

[0002] Organoids are 3D in vitro cell cultures with high structural similarity to their source cells or tissues. They are ideal models for drug testing, replacing patients, thus enabling personalized drug testing and helping doctors and patients improve treatment outcomes. However, currently, the ATP method is mainly used to calculate the survival rate of organoids after drug administration as an indicator of drug efficacy. This method specifically measures cell survival at different drug concentrations and concentrations, generating cell survival curves for each drug at different concentrations, and using the concentration corresponding to 50% survival rate as the efficacy evaluation index. It is evident that the ATP method is an endpoint method; the measurement time affects the measured cell survival rate, which in turn affects the final efficacy evaluation result. Summary of the Invention

[0003] This invention aims to at least partially solve one of the technical problems in the aforementioned technologies. Therefore, one objective of this invention is to propose a deep learning-based organoid quality control method that can circumvent the drawback of the ATP method's sensitivity to measurement time, achieving accurate assessment of organoid developmental quality after drug administration, thereby improving the accuracy of assessing organoid responses to specific drugs.

[0004] The second objective of this invention is to provide a computer-readable storage medium on which a computer program is stored that can circumvent the drawback of the ATP method being sensitive to measurement time, accurately assess the developmental quality of organoids after drug administration, and improve the accuracy of assessing organoid responses to specific drugs.

[0005] To achieve the above objectives, a first aspect of the present invention proposes a deep learning-based organoid quality control method, comprising the following steps:

[0006] Each organoid is selected in the pre-drug and post-drug organoid microscopic images;

[0007] Construct a model that includes convolutional neural networks and multi-task learning neural networks;

[0008] Based on the pre-drug organoid microscopic images and post-drug organoid microscopic images of each organoid selected by the bounding box, the convolutional neural network is trained based on the anchor strategy, so that the convolutional neural network has the ability to identify each organoid from the image.

[0009] Based on the effective features that are discriminative for organoid quality assessment, the multi-task learning neural network is trained so that it has the ability to assess organoid development quality.

[0010] Obtain the evaluation model obtained after training the model.

[0011] According to embodiments of the present invention, a deep learning-based organoid quality control method uses microscopic images of organoids before and after drug administration as training data. Deep learning is performed based on an anchor strategy and a multi-task learning neural network to train an evaluation model capable of identifying each organoid from images and assessing organoid developmental quality. This accurately evaluates the viability of organoids after drug administration, thereby assessing the organoid response to a specific drug. The method of the present invention has the ability to perform repeated tests, avoiding the time-sensitivity of the ATP method, thus improving the accuracy of drug efficacy evaluation.

[0012] In addition, the organoid quality control method based on deep learning proposed in the above embodiments of the present invention may also have the following additional technical features:

[0013] Optionally, selecting each organoid in the pre-drug organoid microscopic image and the post-drug organoid microscopic image includes:

[0014] Original pre-drug organoid microscopic images and original post-drug organoid microscopic images were obtained respectively.

[0015] The original pre-drug organoid microscopic images and the original post-drug organoid microscopic images are preprocessed to obtain corresponding high-throughput pre-drug organoid microscopic images and high-throughput post-drug organoid microscopic images.

[0016] Optionally, the step of training the convolutional neural network based on the pre-drug organoid microscopic images and post-drug organoid microscopic images of each organoid selected by bounding boxes, thereby enabling the convolutional neural network to identify each organoid from the images, includes:

[0017] Preset anchors based on the size of the selected organoids;

[0018] Each cell cluster in the pre-drug organoid microscopic image and the post-drug organoid microscopic image is defined and the bounding box of each cell cluster is obtained.

[0019] The bounding box is encoded relative to the anchor to extract organoid features and obtain the position of the corresponding organoid;

[0020] Based on the obtained organoid locations and the location of each organoid selected by the bounding box, the convolutional neural network is trained using the backpropagation method, enabling the convolutional neural network to recognize each organoid from the image.

[0021] Optionally, the step of training the multi-task learning neural network based on effective features that are discriminative for organoid quality assessment, so that the multi-task learning neural network has the ability to assess organoid development quality, includes:

[0022] Based on a multi-head classifier, effective features with high discriminative power for organoid quality assessment are extracted from each organoid selected by the bounding box;

[0023] Improve the representation capability of the effective features;

[0024] Learning organoid development quality assessment functions based on a contrastive learning strategy;

[0025] Based on the effective features, the enhanced effective features, and the organoid development quality assessment function, the multi-task learning neural network is trained using the backpropagation method, enabling the multi-task learning neural network to assess the development quality of organoids.

[0026] Optionally, the method further includes:

[0027] Microscopic images of organoids were collected at different developmental stages or with different drug concentrations.

[0028] The acquired microscopic images are processed through the evaluation model to obtain the developmental quality of organoids in the corresponding microscopic images.

[0029] The developmental quality of organoid samples in microscopic images is assessed based on the number and developmental quality of organoids in the images.

[0030] To achieve the above objectives, a second aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the above-described deep learning-based organoid quality control method. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating a deep learning-based organoid quality control method according to an embodiment of the present invention.

[0032] Figure 2 This is a schematic diagram illustrating the target bounding box encoding process in a deep learning-based organoid quality control method according to an embodiment of the present invention.

[0033] Figure 3This is a schematic diagram illustrating the process of training a multi-task learning neural network in an organoid quality control method based on deep learning, according to an embodiment of the present invention. Detailed Implementation

[0034] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0035] This invention, based on deep learning technology, captures morphological changes in organoids before and after drug administration using microscopic imaging data, accurately assessing the viability of organoids after drug administration, thereby improving the accuracy of evaluating organoid responses to specific drugs. Because this invention allows for repeated measurements, it overcomes the time-sensitivity limitation of the ATP method.

[0036] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the invention to those skilled in the art.

[0037] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0038] Figure 1 This is a flowchart illustrating a deep learning-based organoid quality control method according to the present invention.

[0039] like Figure 1 As shown in the figure, an organoid quality control method based on deep learning according to an embodiment of the present invention includes:

[0040] S1: Preprocess the image data to improve data clarity.

[0041] In one specific embodiment, this can be achieved through the following steps:

[0042] S101: Select organoids from specific sources and collect organoid microscopic images before and after drug administration, which are recorded as the original pre-drug organoid microscopic images and the original post-drug organoid microscopic images.

[0043] S102: The original organoid microscopic images are stitched together and depth-of-field synthesis to synthesize high-throughput image data, so as to enable clear imaging of organoids with different focal lengths within a single aperture, resulting in high-throughput pre-drug organoid microscopic images and high-throughput post-drug organoid microscopic images.

[0044] S2: Select each organoid in both the pre-drug and post-drug organoid microscopic images, and score each selected organoid based on its morphological characteristics. A score between 0 and 5 is preferred. In this embodiment, the "box" used to select each organoid is defined as the ground truth box.

[0045] S3: Construct a model that includes a convolutional neural network and a multi-task learning neural network;

[0046] S4: Train the convolutional neural network based on the Anchor strategy to enable it to automatically detect each organoid in an image.

[0047] The anchor in the anchor strategy refers to the position information of several predefined bounding boxes, which consists of four parameters [x anchor y anchor w anchor h anchor The parameters represent the x-axis coordinate, y-axis coordinate, width, and height of the anchor, respectively. They can also be simply understood as pre-defined reference boxes of different sizes and aspect ratios on the image. In this embodiment, the desired training of the convolutional neural network is to achieve the following: First, feature maps of 4*4, 2*2, and 1*1 sizes are generated by downsampling the microscopic image input to the neural network, such as by 64, 128, or 256 times. Then, a preset number of anchors of different sizes are set on each feature map, for example, three boxes of different sizes and aspect ratios. Next, it is determined whether each anchor contains (or significantly overlaps with) an organoid, and the offset of the organoid relative to the center point of the anchor and its aspect ratio. Finally, the probability that each anchor believes it contains an organoid, the offset of the organoid's center point from the anchor's own center point, and its aspect ratio relative to the anchor's width and height are output. Since the anchor's position is fixed, the organoid's position is easily calculated.

[0048] In one specific embodiment, the above-mentioned S4 can be implemented by the following steps:

[0049] S401: Based on the size of the selected organoid, preset the anchor size and aspect ratio;

[0050] S402: Define the bounding boxes of each cell cluster in the pre-drug organoid microscopic image and the post-drug organoid microscopic image respectively.

[0051] S403: Encode the bounding box relative to the anchor to obtain candidate boxes. If the IoU between the candidate box and a ground truth box is large, it is considered a positive sample and marked as a predicted box; otherwise, it is a negative sample.

[0052] like Figure 2 As shown, corresponding to the organoids in the figure, the white box in the figure is the Anchor, which consists of four parameters [x anchor y anchor w anchor h anchor [σ] represents the x-axis coordinate, y-axis coordinate, width, and height of the anchor, respectively; the algorithm model predicts candidate boxes equal to the number of anchors, i.e., the black boxes in the figure. The encoding format of the candidate boxes is [σ]. x , σ y , σ w , σ h ] indicates that the actual position of the candidate box corresponding to each anchor is [x anchor +δ x y anchor +δ y w anchor +δ w h anchor +δ_h]. That is Figure 2 In the diagram, the black boxes represent candidate boxes corresponding to the white boxes (Anchors). If the algorithm performs perfectly, the candidate boxes will perfectly overlap with the ground truth labeled boxes. This is also the training objective of this step.

[0053] S404: Extract organoid features from positive samples;

[0054] Alternatively, it can be done using the formula: h i =F(x) i Organoid feature extraction is performed, where h i Let F represent the features of each organoid, and let x represent the feature extraction module. i Represents an organoid image block.

[0055] S405: Output the target location of the organoid.

[0056] Alternatively, it can be done using the formula: [p i , t i ]=D(F(x i Obtain the location of the organoid, where [p i , t i] represents the predicted probabilities and locations of all organoids in the image, and D represents the mapping from features to locations, consisting of a series of convolutional layers.

[0057] S406: Based on the obtained organoid locations and the location of each organoid selected by the bounding box, the convolutional neural network is trained using the backpropagation method, so that the convolutional neural network has the ability to identify each organoid from the image.

[0058] In one specific embodiment, the loss value can be calculated by comparing it with the label value (i.e., the actual location of the selected organoid), and the convolutional neural network in the model can be trained by backpropagation using the following formula.

[0059] The formula is: in, Representing image x i The confidence and position of the j-th organoid in the loss function, where BCE represents the binary cross-entropy function and λ represents the weighting coefficients of the two terms of the loss function.

[0060] S5: Train a neural network based on multi-task learning to enable it to assess the quality of organoid development.

[0061] In one specific embodiment, please refer to Figure 3 The above steps can be achieved through the following process:

[0062] S501: Based on a multi-head classifier, extract effective features with high discriminative power for organoid quality assessment from each selected organoid.

[0063] Optionally, via formula Extraction is performed. Among them, Represents the classification loss function. C represents the annotation of the j-th expert on the i-th organoid cell cluster. j Let G(x) represent the j-th classifier. i ) represents a feature extractor.

[0064] S502: Enhance the representation capability of the effective features.

[0065] In a specific example, the representational power of organoid features is improved based on neighborhood constraints. Specifically, this is achieved through the formula:

[0066]

[0067]

[0068] Improve the representation capability of effective features. Among them, Represents the clustering loss. p represents the prediction value of the j-th classifier for the i-th cell cluster. uv h represents the similarity between the u-th organoid and the v-th organoid. u Represents the characteristics of the u-th organoid. Z represents the transpose of the feature representation of the v-th organoid. u This represents the normalization coefficient.

[0069] S503: Learning organoid development quality assessment functions based on contrastive learning strategies;

[0070] In a specific instance, it can be achieved using the following formula:

[0071]

[0072]

[0073] in, Let S represent the contrastive loss function, and S represent the exponential function, which returns a specific value based on given conditions. u x v ) represents organoid sample pairs, Let represent the annotation value of the j-th expert for the u-th organoid. Let represent the annotation value of the j-th expert for the v-th organoid. This represents any expert's annotation of sample v. Let represent any expert's label for sample u, and M represent the scoring function.

[0074] S504: Based on the effective features, the enhanced effective features, and the organoid development quality assessment function, the multi-task learning neural network is trained using the backpropagation method, enabling the multi-task learning neural network to assess organoid development quality. See details for implementation. Figure 3 Model.

[0075] In a specific instance, the loss value can be calculated by comparing it with the label value (i.e., the actual location of the selected organoid), and the multi-task learning neural network in the model can be trained by backpropagation using the following formula.

[0076] The formula is: Where α represents the weighting coefficient and β represents the weighting coefficient.

[0077] S6: Obtain the trained organoid development quality assessment model.

[0078] Based on this organoid development quality assessment model, it is possible to capture morphological changes in organoids, which is beneficial for a more accurate assessment of the drug efficacy of organoids.

[0079] In one specific embodiment, the method further includes:

[0080] S7: The developmental quality of organoids is assessed using the trained organoid developmental quality assessment model.

[0081] In a specific instance, the evaluation process includes:

[0082] S701: Add the drug to be evaluated into a culture dish containing organoids, and collect microscopic images of organoids at different drug concentrations or different developmental times.

[0083] S702: The acquired microscopic images are processed through the evaluation model to predict the developmental quality of organoids in the input microscopic images.

[0084] S703: Based on the number of organoids in the microscopic image and the developmental quality of each organoid, comprehensively evaluate the developmental quality of the organoid samples in the microscopic image.

[0085] In summary, the deep learning-based organoid quality control method provided in this embodiment uses microscopic images of organoids before and after drug administration as training data. Deep learning is performed based on an anchor strategy and a multi-task learning neural network to train an evaluation model capable of identifying each organoid from images and assessing organoid development quality. This accurately evaluates the viability of organoids after drug administration, thereby evaluating the organoid response to a specific drug. The method of this invention has the ability to perform repeated tests, avoiding the time-sensitivity of the ATP method, thus improving the accuracy of drug efficacy evaluation.

[0086] In addition, this embodiment of the invention also proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, can implement the above-mentioned deep learning-based organoid quality control method.

[0087] According to embodiments of the present invention, a computer-readable storage medium storing an evaluation model enables the implementation of the aforementioned deep learning-based organoid quality control method when executed by a processor. Thus, based on the evaluation model, the viability value of organoids after drug administration can be accurately assessed, thereby evaluating the organoid response to a specific drug. Because it possesses the ability to perform repeated tests, it avoids the drawback of the ATP method's sensitivity to measurement time, thereby improving the accuracy of drug efficacy evaluation.

[0088] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0089] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0090] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0091] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0092] It should be noted that any reference signs placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0093] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0094] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0095] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0096] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0097] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0098] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms should not be construed as necessarily referring to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0099] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for quality control of organoids based on deep learning, characterized in that, Includes the following steps: Each organoid is selected in the pre-drug and post-drug organoid microscopic images; Construct a model that includes convolutional neural networks and multi-task learning neural networks; Based on the pre-drug organoid microscopic images and post-drug organoid microscopic images of each organoid selected by the bounding box, the convolutional neural network is trained based on the anchor strategy, so that the convolutional neural network has the ability to identify each organoid from the image. Based on the effective features that are discriminative for organoid quality assessment, the multi-task learning neural network is trained so that it has the ability to assess organoid development quality. Obtain the evaluation model obtained after training the model; The process involves training a multi-task learning neural network based on effective features that are discriminative for organoid quality assessment, thereby enabling the network to evaluate organoid development quality. This includes: extracting highly discriminative effective features for organoid quality assessment from each selected organoid using a multi-head classifier; enhancing the representational power of the effective features; learning an organoid development quality assessment function based on a contrastive learning strategy; and training the multi-task learning neural network using backpropagation based on the effective features, the enhanced effective features, and the organoid development quality assessment function, thus enabling the network to evaluate organoid development quality. Among them, the effective features with high discriminative power for organoid quality assessment are extracted according to the following formula: in, Represents the classification loss function. This represents the annotation of the j-th expert on the i-th organoid cell cluster. This represents the j-th classifier. Indicates a feature extractor; The representational power of effective features is improved according to the following formula: in, Represents the clustering loss. This represents the prediction value of the j-th classifier for the i-th cell cluster. This represents the similarity between the u-th organoid and the v-th organoid. Represents the characteristics of the u-th organoid. This represents the transpose of the feature representation of the v-th organoid. Represents the normalization coefficient; The organoid development quality assessment function is learned according to the following formula: in, This represents the contrastive loss function. This represents an exponential function that returns a specific value based on given conditions. ) represents organoid sample pairs, Let represent the annotation value of the j-th expert for the u-th organoid. Let represent the annotation value of the j-th expert for the v-th organoid. This represents any expert's annotation of sample v. Let represent any expert's label for sample u, and M represent the scoring function.

2. The deep learning-based organoid quality control method as described in claim 1, characterized in that, The process of selecting each organoid in pre-drug and post-drug organoid microscopic images includes: Original pre-drug organoid microscopic images and original post-drug organoid microscopic images were obtained respectively. The original pre-drug organoid microscopic images and the original post-drug organoid microscopic images are preprocessed to obtain corresponding high-throughput pre-drug organoid microscopic images and high-throughput post-drug organoid microscopic images.

3. The organoid quality control method based on deep learning as described in claim 1, characterized in that, The process involves selecting pre-drug and post-drug organoid microscopic images for each organoid, and training the convolutional neural network based on an anchor strategy. This enables the convolutional neural network to identify each organoid from the images, including: Preset anchors based on the size of the selected organoids; Each cell cluster in the pre-drug organoid microscopic image and the post-drug organoid microscopic image is defined and the bounding box of each cell cluster is obtained. The bounding box is encoded relative to the anchor to extract organoid features and obtain the position of the corresponding organoid; Based on the obtained organoid locations and the location of each organoid selected by the bounding box, the convolutional neural network is trained using the backpropagation method, enabling the convolutional neural network to recognize each organoid from the image.

4. The organoid quality control method based on deep learning as described in claim 1, characterized in that, The method further includes: Microscopic images of organoids were collected at different developmental stages or with different drug concentrations. The acquired microscopic images are processed through the evaluation model to obtain the developmental quality of organoids in the corresponding microscopic images. The developmental quality of organoid samples in microscopic images is assessed based on the number and developmental quality of organoids in the images.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it can implement the deep learning-based organoid quality control method described in any one of claims 1-4.

Images